Picomolar inhibition of cholera toxin by a pentavalent ganglioside GM 1 os-calix [ 5 ] arene †

Cholera toxin (CT), the causative agent of cholera, displays a pentavalent binding domain that targets the oligosaccharide of ganglioside GM1 (GM1os) on the periphery of human abdominal epithelial cells. Here, we report the first GM1os-based CT inhibitor that matches the valency of the CT binding domain (CTB). This pentavalent inhibitor contains five GM1os moieties linked to a calix[5]arene scaffold. When evaluated by an inhibition assay, it achieved a picomolar inhibition potency (IC50 = 450 pM) for CTB. This represents a significant multivalency effect, with a relative inhibitory potency of 100 000 compared to a monovalent GM1os derivative, making GM1os-calix[5]arene one of the most potent known CTB inhibitors.


Introduction
Cholera still represents a serious health problem in areas of the developing world where there is a lack of clean water and proper sanitation. In 2012, the World Health Organization estimated that annually 3-5 million cholera cases occur that result in more than 100 000 deaths. 1 Although several treatments exist for cholera, 2 resistance development and mutations in the causative pathogen mean that efforts made to better understand the disease pathogenesis and develop new treatments are crucial. 3,4 The symptoms of cholera are caused by cholera toxin (CT), which is produced by the Vibrio cholerae bacterium. CT is a member of the AB5 toxin family that contains a pentameric binding domain (CTB) for recognition and binding to cell surfaces. 5 The natural target ligand for CTB is the glycosphingolipid ganglioside GM1, on cellular membranes of the infected hosts' intestinal epithelial surface. CTB can bind five GM1 saccharide epitopes simultaneously with the terminal Gal-and the Neu5Ac carbohydrate units of the ganglioside as the major contributors to the binding. 6,7 The adhesion of CTB to ganglioside GM1 on cell surfaces is the prerequisite for endocytosis of the toxic enzymatically active A subunit of CT, and the ensuing severe clinical symptoms. 8 One avenue in cholera research is to study the binding of CTB to GM1 and to develop CTB inhibitors that might prevent CT from binding the hosts' cell surface and thereby also the development of cholera. Here, we present the second of two examples of a pentavalent GM1os-based inhibitor for CTB, GM1os-calix [5]arene (1; Fig. 1). In the previous paper in this issue, we also reported on pentavalent inhibitors of CTB based on a GM1os-presenting corannulene scaffold. In the past, several studies have focused on the development of multivalent glycosylated inhibitors for CTB based on ganglioside GM1. 9 It is noteworthy that in none of these studies, inhibitors were investigated with a pentavalent structure that matches the pentavalent structure of CTB. On the other hand, pentavalent  10,11 and cyclic peptides 12 have been described as CTB inhibitors, but those contained only the much simpler galactose epitope likely to get around the difficulty to obtain sufficient tailor-made GM1os. Therefore, also GM1 mimics have been used, e.g. Thompson and Schengrund described poly( propylene) imine dendrimers that present the Galβ1-3GalNAcβ1-4(Neu5Acα2-3)Galβ-epitope of GM1, 13 with IC 50 values for the tetravalent and octavalent dendrimers of 7 and 3 nM, respectively. Bernardi et al. published a series of GM1mimics ( pseudo-GM1), in which the residues in the GM1os that are not essential for binding were replaced by a conformationally restricted cyclohexane-diol and the Neu5Ac-unit was substituted by various α-hydroxy acids. 14,15 When attached to multivalent dendritic structures, 16 the relative inhibitory potency (RIP) values per mimic unit of the tetravalent and octavalent inhibitors were 111 and 55, respectively. Interestingly, when the same mimic was linked to a divalent calix [4]arene scaffold, 17 a 4000-fold enhancement in binding efficiency was achieved compared to the monovalent pseudo-GM1. These data suggested to us that the calixarene macrocycle, from which the binding inhibitors are projected, could be a promising multivalent scaffold [18][19][20] to design CTB inhibitors with improved efficiency. In collaboration with the group of Pieters, we previously published divalent, tetravalent, and octavalent dendritic structures decorated with GM1os. 21,22 For the octavalent compound the unprecedentedly low IC 50 value of 50 pM was observed with an RIP of 17 500 per arm compared to its monovalent counterpart. However, its mismatched valency compared to CTB prompted us to investigate a pentavalent scaffold as core structure that when decorated with GM1os has the potential to form 1 : 1 inhibitor-CTB complexes. The current paper presents the convergent synthesis of the first, water-soluble, pentavalent CTB inhibitor (1), which was made by coupling five GM1os units to a calix [5]arene scaffold.

Results and discussion
We designed a 5-fold symmetric calix [5]arene as a pentavalent scaffold structure. This calix [5]arene ( Fig. 1) presents small methoxy groups at the lower rim, which confer a high conformational flexibility to the macrocyclic structure. 23 The upper rim of the calixarene inhibitor is decorated with the GM1 pentasaccharide separated from the macrocyclic core by appropriate linkers. Fan and coworkers 12 have demonstrated that an optimal linker length is vital for the potency of a synthetic multivalent inhibitor. For the calix [5]arene, described here, a 31 atom-containing linker was chosen. This should allow the simultaneous interaction of the five GM1os units with the five B-subunits of a single toxin. 5 The route towards our target (1) started with the synthesis of the pentavalent scaffold that began with the preparation of the known p-tert-butyl-calix [5]arene. 24 This product was converted into p-H-calix [5]arene 2 25 by following literature procedures. Next, penta-aldehyde 3 was obtained in 57% yield from 2 by exploiting the Duff formylation reaction. 26,27 Compound 3 was subsequently methylated at the lower rim by using CH 3 I and K 2 CO 3 in acetonitrile affording the pentamethoxy-calix [5]arene 4 in 68% yield (Scheme 1).
Oxidation of 4 with NaClO 2 and NH 2 SO 3 H produced the penta-carboxylic acid 5. The unsymmetrically substituted azido-penta-(ethyleneglycol)-amine 7 was synthesized from hexa-ethylene glycol by ditosylation, substitution to the diazide, and finally selective Staudinger reduction of one azide. 28,29 Initial attempts to condense amine 7 with the carboxylic acids Scheme 1 Synthesis of the penta-azido-calix [5]arene (8)  Organic & Biomolecular Chemistry Paper in 5 using HBTU resulted in difficult purification and low yields (∼20%) of 8. However, when this spacer (7) was attached to calix [5]arene 5 via penta-acyl chloride intermediate 6 it provided penta-azido-calix [5]arene 8 in a 44% yield. With the calix [5]arene (8) scaffold in hand we proceeded to the next stage, attaching five GM1 oligosaccharides. We chose the copper-catalyzed azide-alkyne cycloaddition (CuAAC) reaction to achieve this, which meant a GM1os derivative with a terminal alkyne was required. This C 11 -alkyne-terminated GM1os 9 (Scheme 2) was made via a chemo-enzymatic procedure previously reported by us, 30 which allowed the production of 9 on gram scale. Compound 9 was subsequently "clicked" to scaffold 8 under standard CuAAC conditions in H 2 O while exposed to microwave irradiation to successfully provide our crude target inhibitor 1. With our target pentavalent GM1os-calix [5]arene 1 in hand, in order to properly assess the role of the GM1os in inhibitor 1, we also set out to synthesize derivatives of 1 containing fragments of the GM1os to use for comparison in the biological assays. The first of these was penta-GM2os-calix [5]arene (11) that lacks the terminal galactose epitope compared to the GM1os. We synthesized 11 using the same CuAAC reaction conditions from scaffold 8 and a chemo-enzymatically produced alkyne-terminated C 11 -linked GM2os sugar (10).
With both our target 1 and its derivative 11 in hand as crude products, we investigated what purification method would be suitable for these large complex molecules. An initial purification by size exclusion chromatography (SEC) efficiently removed an excess of alkyne-terminated GM1os 9 and GM2os 10, respectively. However, the crude products both also contained a minor amount of tetravalent byproducts as could be clearly seen with mass spectrometry and their separation proved to be quite challenging. Initial attempts to separate these by aqueous HPLC GPC failed, but after extensive optimization, reversed phase HPLC purification proved the most successful for this final purification step (see experimental section for details). Fig. 2 shows a typical HPLC chromatogram for the separation of both the GM1os-and GM2os-calix[5]arenes. Despite attempts to improve the moderate resolution, the purification remained quite complicated because of long elution times (up to 40 minutes per run) and high affinity of the products with the column material. Collection of small fractions in a specific retention time window over multiple HPLC injections and subsequent lyophilization resulted indeed in pure pentavalent GM1os-and GM2os-calix [5]arenes (1 and 11) as shown by HR-MS and NMR analyses. The other impure fractions that also contained 1 or 11 were collected, pooled, lyophilized and re-injected to achieve optimal yields. Attempts were made to isolate the tetravalent byproducts, but insufficient amounts could be obtained for further analyses.
Besides the GM2os containing calixarene (11), we also prepared two further derivatives of 1 that contained fragments of GM1os, a pentavalent β-galactoside-(16) and β-lactoside-calix- [5]arene (17) (Scheme 3). These more simple carbohydrates enabled a modified synthesis procedure that circumvented the potentially challenging HPLC purification, as encountered for compounds 1 and 11. The coupling was also performed by employing the microwave-assisted CuAAC reaction on penta-azido-calixarene scaffold 8, but instead of using the deprotected carbohydrates, acetyl-protected galactoside 12, and lactoside 13 were reacted. The resulting products could now be purified by normal phase silica gel chromatograpy. The acetylprotected 14 and 15 were deprotected by employing standard Zemplén 31 conditions to obtain galactoside-calix [5]arene 16, and lactoside-calix [5]arene 17, respectively, which did not require further purification after work-up.
Finally, in order to properly determine the multivalency effect of the interaction of 1 with CTB in our biological assays, we also synthesized the monovalent GM1os derivative 20 (Scheme 4). This was achieved by first in situ generation of the acyl chloride of commercially available 4-methoxybenzoic acid with oxalyl chloride and, subsequently, reacting this with amino-azide 7, yielding azide 19 in 20% over two steps. Again, by employing the microwave-assisted CuAAC reaction on alkyne-terminated 9 and azide 19, GM1os-monomer 20 was obtained in a reasonable yield (49%).
The inhibitory potency of the four pentavalent compounds (1, 11, 16, and 17) was determined by ELISA experiments. In the assays, the ability of 1, 11, 16 and 17 to inhibit the binding of HRP-labeled CTB was measured in competition with the natural ligand ganglioside GM1, which was adsorbed to the well surface of the ELISA plate. GM1os-calix [5]arene 1 showed a high inhibition potency, i.e., a very low IC 50 value of 450 pM (Fig. 3, Table 1). Comparing the IC 50 value (44 μM) of the monovalent control compound (20) to that of 1 revealed a 100thousand increase in inhibitory potency, and an RIP of 20thousand per arm. Pentavalent inhibitors based on a more rigid corannulene scaffold, which we also report in this issue, inhibited CTB in the nanomolar range. 32 Assay results for the GM2os-calix [5]arene 11 confirmed the importance of using GM1os. Lacking only the terminal galactose compared to 1, it produced an IC 50 of 9 μM, which is 20-thousand fold worse compared to 1. The galactose-terminated (16) and lactose-terminated (17) calix [5]arenes displayed a higher inhibition concentration than their solubility in the assay medium, and their IC 50 could therefore only be determined as being >1 mM (Table 1).

GM1-calix[5]arene (1)
Calix [5]arene 8 (15.7 mg, 6.93 μmol) was dissolved in 0.5 mL of CH 3 OH in a microwave tube. Then the GM1os derivative 9 (59.8 mg, 52.1 μmol), previously synthesized by chemo-enzymatic procedures, 30  GM1os-calix [5]arene GM2os-calix [5]arene GM1os-monomer; Fitted curves of the experimental ELISA inhibition data. For details of the inhibition assays see experimental section.  13  5,11,17,23,29-Pentakis[(17-azide-3,6,9,12,15-pentaoxaheptadecane-1-amino)carbonyl]-31,32,33,34,35-pentamethoxy-calix- [5]arene (8) In a round-bottomed flask, 0.12 g of calix [5]arene 5 (0.14 mmol) and 0.51 mL of oxalyl chloride (5.82 mmol) were solubilized in 15 mL of dry CH 2 Cl 2 under a nitrogen atmosphere. The solution was stirred for 18 h at room temperature and then the solvent evaporated to dryness. The residual compound 6 was dissolved again in 5 mL of dry CH 2 Cl 2 and then added dropwise to a solution of amine compound 7 (0.29 g, 0.87 mmol) and NEt 3 (0.12 mL, 0.87 mmol) in 5 mL of dry CH 2 Cl 2 . The mixture was stirred for 20 h at room temperature under a nitrogen atmosphere. The mixture was then washed with 1 M HCl, an aqueous solution of Na 2 CO 3 and water till neutral pH was reached. The solvent was removed under vacuum and the crude purified by flash chromatography (eluent: CHCl 3 -CH 3 OH 95 : 5) to give the product 8 as a yellow oil. Yield: 44%. 1  Galactosylcalix [5]arene (16) The peracetylated-galactosylcalix [5]arene 14 (45.0 mg, 9.46 μmol) was dissolved in 5 mL of CH 3 OH, drops of a freshly prepared MeONa in methanol solution were added till pH 8-9. The mixture was stirred at room temperature for 4 h. The progress of the reaction was monitored via ESI-MS analysis. Amberlite resin IR 120/H + was subsequently added to quench the reaction, and the mixture was gently stirred for 30 min. until neutral pH was reached. The resin was then filtered off and the solvent removed under vacuum to give pure product 16 as a yellow oil. Yield. 90%. 1  Lactosylcalix [5]arene (17) The peracetylated-lactosylcalix [5]arene 15 (42.0 mg, 6.8 μmol) was dissolved in 5 mL of CH 3 OH, and drops of a freshly prepared methanol solution of MeONa were added till pH 8-9. The mixture was stirred at room temperature for 18 h. The progress of the reaction was monitored via ESI-MS analysis. Amberlite resin IR 120/H + was subsequently added for quenching, and the mixture was gently stirred for 30 min till neutral pH. The resin was then filtered off and the solvent removed under vacuum to give pure product 17 as a yellow oil. Yield: 72%. 1